Re: Improve internal chord structure

[David Kastrup]

address@hidden writes:

>>>> I hope to send something more specifically about chords after some rest.
>>>
>>> Here are my 2cents:
>>> Defining the chords by a list of notes for jazz tunes can't  ever
>>> cover all sorts of voicings.
>>
>> Are you confusing "pitch" and "note" here?  Notes are a LilyPond data
>> structure that have more properties than just a pitch.  Several are
>> already used for disambiguating chords entered in chord mode.
>
> Very likely, i don't know much of the lilypond internals.
>
>>> Fortunately, this isn't at all needed for typesetting e.g. a
>>> jazz-tune.  It is sufficient to only have the root note, e.g. C from
>>> C:m7, in a musical/transposeable context, the rest of the chord(here
>>> m7) could just be represented as text as-it-is.
>>>
>>> Maybe, from this point of view, there could be an alternative method
>>> for typestting chords be implemented?
>>
>> I don't see what you are trying to achieve here.
>
> English is not my native language, but i try:
> I wanted to point out, that the idea to get e.g.
> c' e' g' a' from the chord c:6 is not really correct/complete and not
> of much use, for various reasons:
> -unclear in which octave
> -unclear which voicing
> -Not all pitches needed to be there(may be too much/dense) to make the
> chord clear
> -c' e' g' a' could be interpreded as chord a:m7

Please take a look at

mus = \chordmode { c2:6 a:m7/c }

<<
  \new ChordNames \mus
  \new Staff \mus
>>

\void \displayMusic \mus

You'll find that the same notes can already be distinguished as either
chord/inversion.

One question certainly is whether the information we use for that is a
good fit and whether it should be easier to create this kind of
information outside of \chordmode .

But it's not like there is nothing there yet.

> In a reallife jazz context, it's up to the player/band what to do with
> chords. Lilypond is not a playalong-generator, but lilypond should
> typeset chords. Lilypond can do this in most cases, but in my opinion
> the mapping to single pitches/notenames is an unneeded, complicated
> and errorprone overhead.

If you want to output Midi, you need pitches anyway.

> So, reducing the lilypond chord-handling to a singel pitch/root-note +
> text/chord-description would be sufficient to typeset.
> This way, all sorts of chord-types. e.g. c7/b9/#9/b5/whatever are
> possible, without worrying about implementing all possibilities how a
> chord could be interpreted.

Text does not transpose very well.  Various instruments (guitar, banjo,
accordion) have various ways of picking voicings that don't correspond
with the "canonical" description but still have some representation in
tablature/score.

One current problem with chord naming and entry is that, for example,
guitar voicings are hardwired into FretBoard contexts and you can't
actually get them out of there (for example, for putting them in
tablature).  It's also hard to create descriptive chord names when you
start with a score.

There are a number of problems where the current implementation and its
representation makes things harder to bounce around.

-- 
David Kastrup